Modulation techniques of chemical sensors for analyte identification

Abstract

The modulation of chemical sensors and the signal processing of the resultant dynamic signals are described as an alternative to the traditional techniques of analyte identification in artificial olfaction systems. Experimental apparatus were developed for the controlled and repeatable modulation of these sensors, as well as data acquisition and analysis. The temperature modulation of SnO2 sensors and signal analysis by the discrete wavelet transform (DWT) was found to produce adequate discrimination between three volatile organic vapours (benzene, ethanol, THF), but lacked long term stability due to sensor drift. A number of polymer coated TSM sensors were evaluated for factors such as sensitivity, linearity and long-term stability. Cellulose acetate and polyethylene-oxide (PEO) coated sensors showed the best overall properties. These TSM sensors were then modulated by a concentration modulation technique and the dynamic signals analysed by the DWT. This was found to produce good clustering for the selected analytes, but failed when the analyte concentration was varied. A novel method of concentration-series modulation was developed, which exposed the TSM sensor to a modulation pulse over a range of analyte concentrations. Two dynamic parameters, the rise time and the dynamic frequency, were extracted from the transient part of the sensor signal and combined with the steady state frequency shift of the sensor. These three parameters were used to produce a description of the analyte in three-dimensional (3-D) spaces. This data could be used as the calibration data for an analyte in a database. This method was shown to be successful in the identification of these three analytes from a group of five unknowns, with an absolute error of < 6% in the variation between unknown and calibration data. In addition, a good estimate of the analyte concentration could also be obtained by this method. The method was also applied to a much more complex problem in the classification and identification of three commercial brands of herbal tea, and a successful identification of all three brands was shown to be possible with this method.